Performance with Connect by

Question

All,
Please help with tuning the query. It takes a long
delete from XXN_T2 A
where A.PROGRAM = '00HMS'
and not exists
(select AA.CODE
from BDNDETL AA
where AA.BREAKFILE = 'HMSWBS'
and trim(A.CNTRL_ACCT) = trim(AA.CODE)
start with CODE in (select BB.CODE
from BDNDETL BB
where BB.BREAKFILE = 'HMSWBS'
and AA.CODE = BB.CODE
and AA.TAG = BB.TAG
and BB.BDN_LEVEL = 2)
and BREAKFILE = 'HMSWBS'
connect by prior CODE = parent
and BREAKFILE = 'HMSWBS'
)

Answer 1

Hi,
Instead of doing a Top-Down CONNECT BY query, where you start with certain ancestors and then see if they have certain descendants, make the EXISTS sub-query a Bottom-Up query, where you start with the descendants and then see if they have certain ancestors.
Edited by: Frank Kulash on Jun 10, 2009 5:12 PM
What I mean is something like:
AND     NOT EXISTS
     (     SELECT     1
                 FROM     bdndetl          aa
          WHERE     aa.breakfile          = 'HMSWBS'
          AND     EXISTS (       SELECT     1
                                      FROM     bdndetl     bb
                           WHERE     bb.breakfile = 'HMSWBS'
                           AND     aa.code          = bb.code
                           AND     aa.tag          = bb.tag
                           AND     bb.bdn_level = 2
          START WITH     TRIM (a.cntrl_acct)     = TRIM (aa.code)
          AND               breakfile           = 'HMSWBS'
          CONNECT BY        code           = PRIOR parent
          AND          breakfile     = 'HMSWBS'
     )Without some sample data and the desired results (i.e., what you want in table a after the DELETE), I can't test it.
I made the sub-query on table bb an EXISTS sub-query, also. This is independent of the Bottom-Up query: you can try either one without the other.